Anotaciones de enfermería

The NICE Methods Guide refers to a number of types of information need in its guidance on model-based cost-effectiveness analyses. The guide also points to the need to identify different study designs to address different types of information need and to use both research and non-research based information.

A content analysis of the evidence cited in typical models of cost-effectiveness identified fourteen types of information drawn from seven types of sources.8 (see Table 2). This included the citing of sources of evidence to support the full scope of modelling activities. In terms of populating model parameters, major types of input have been defined as treatment effects, costs, utilities, baseline risk of clinical events11,23 and adverse events.19 The main types of sources from which evidence to populate model parameters is drawn has been classified as evidence syntheses (including secondary economic analyses), RCTs, observational studies (including both longitudinal and cross-sectional studies), routine data sources (including registries and routine and administrative data sources), references sources (including drug formularies) and expert judgment.8 The range of types and sources of evidence used in models has implications for the retrieval of information.

41

Table 2: Classification of evidence used in models of cost-effectiveness by type, source and use of evidence (Paisley, 201021)

Types of information Types of source of evidence Uses of evidence within model

Adverse effects Evidence synthesis Design and specification of model

framework

Compliance Expert judgment Model validation

Current practice Methodological theory and

empirical evidence Modelling and analytical approach

Epidemiology Observational research Population of model parameters

Modelling methods RCT (clinical and economic) Sensitivity and uncertainty analysis

Natural history Reference sources

Patient preferences Routine data sources

Prescribing rates

Prognosis

Resource use

Results and methods from other models

Clinical outcomes

Unit costs

Utilities

General biomedical bibliographic databases, including Medline and Embase remain important sources. The practice of using validated filters to restrict search results to RCTs is well established. Filters also exist for other types of study design. However, many of these have been designed pragmatically and few have been widely validated. This should be borne in mind when performing searches aimed at achieving high sensitivity. Despite this caveat, filters for study types other than RCTs are widely used and are an important resource in identifying the range of information required for models. The Information Specialists’ Sub- Group of InterTASC, the network of academic centres undertaking health technology assessments for NICE, have developed an extensive resource of critically appraised search

42

Working Group provides advice on searching for economic evaluations and on utilities

evidence.19,25 Another in the series of NICE DSU Technical Support Documents focuses on

the identification of utilities evidence.26 Research on the retrieval of evidence on adverse events is currently being undertaken at the University of York.27 These sources provide advice both on searching general biomedical databases and on specialist resources specific to their topic areas.

Compilations of resources bringing together routine data sources such as registries and administrative data sources and reference sources such as drug formularies and unit costs have not been developed. Part of the value of this type of information is its relevance to the context of the decision-making process, such as geographical context. For example the British National Formulary (BNF), the Office for National Statistics (ONS) and the NHS Reference Costs provide relevant information in the context of NICE decision-making but do not carry the same authority in all decision-making jurisdictions. As such it would be difficult to develop a generic, comprehensive resource useful to all HTA decision-making systems. However this type of information source is difficult to locate systematically. The compilation of resources relevant to specific, local decision-making processes is an important area for further development. Many of these resources are already mentioned in the NICE Methods Guide. Further methods of identifying this type of information include the use of expert advisors in identifying important sources in their clinical area and brief exploratory searches of the internet and of bibliographic databases to identify leads or ‘proximal cues’ that can be followed to potentially relevant sources.

The role of expert judgment as a source of evidence in decision-analytic models is more complex than in systematic reviews. As a source of estimates with which to populate model parameters, the expert judgment of clinicians is placed at the bottom of the evidence hierarchy, as in systematic reviews.23 However, in the absence of relevant of evidence, the elicitation of expert information is preferable to an unqualified assumption. Beyond the population of model parameters, the role of expert judgment in interpreting the available evidence and in assessing the face validity or credibility of the model as an acceptable representation of the decision problem is recognised as an important form of validation.9 In terms of retrieving or obtaining information from experts, the potential for bias is high and the guiding principles pertaining to the more familiar methods of information retrieval remain in place. Formal methods of elicitation, including Bayesian elicitation methods, qualitative

43

operational research techniques22,28 and consensus methods29 have been used in the context of decision-analytic models. Other forms of good practice include the use of a number of experts in order to capture a range of opinion and variation and taking precautions to ensure that the full range of clinical expertise relevant to the decision problem is represented (see Table 1).

In the context of different types of information (e.g. epidemiology, clinical effectiveness, resource use) it is perhaps worth a further consideration of the scope of evidence required to inform a model. Systematic review search methods focus on the identification of direct evidence for inclusion in a review. That is, the search aims to identify studies that seek to address the same question as the review. In order to do so, the scope of the search strategy is structured according to the specification of the PICO question. Whilst a model might address the same PICO question as a systematic review, it does so within a broader analytic framework that aims to reflect the complexity of the decision problem. In bringing together evidence within this broader framework, models consider a wide range of evidence both directly and indirectly related to the specified PICO question. In particular, models require information in order to assess the impact of a technology over the course of a disease. As such, the scope of relevant evidence is dictated by the definition of health states and events within the model rather than by the underlying PICO question. For example, in terms of evidence on effectiveness, costs and resource use, utilities, adverse events and baseline risk of events, the scope of information required is not reflected explicitly in the PICO question but is dictated by the treatment options, management of the disease and population characteristics at all other stages of the disease pathway included in the model.